Comment on A first-person 3D adaptation of Zork I

• dbtng@sh.itjust.works ⁨2⁩ ⁨days⁩ ago

  Well in this game, there was the chair (see my other comment) that no human would sit in.
  There are other tells that process that doesn’t understand the continuity of solid objects created the game, not a person.

  • You can walk through most of the walls and half of the scenery. Walls don’t normally work like that.
  • The explorer skeleton in the twisty maze room simply is not there. He should be sitting beside the key that’s nearly fukin impossible to find because there’s no skeleton next to it, despite the room description saying he’s there.
  • The paddle for the inflatable raft is in two pieces, the blade sits about a foot from the shaft it should be attached to.
  • The diamond mill output is … nothing. It’s just a spot on the floor, uncomfortably far away and disconnected from the mill.
  • The area where you have to deal with coal gas is just a complete fail, the AI did not understand and just made some crap.

  .
  There were two other questionable bits where I don’t remember the original game, but it seems the AI flubbed. The 4 button dam controls were wonky: stop, start, bust the pipe, and what? And the cyclops room has a door you never get to open, no info about it.
  All together, these seem like things no human would get wrong. Like, maybe a human would make a shitty chair, but they would almost certainly ensure that all the pieces fit together before they published the game.

  source

  • General_Effort@lemmy.world ⁨1⁩ ⁨day⁩ ago

    That’s some great testing/feedback. I hope Ethan sees that.

    A human sees the errors immediately but fixing still takes time. One would prioritize getting a working prototype out before polishing away such issues. Rationally, it’s more important to get feedback on the mechanics and such than making sure the models and collisions are good.

    I believe you are right that a human wouldn’t make at least some of these errors. But I think one needs some experience with 3D modelling software or alpha stage games to really be confident about what errors humans plausibly make.

    source
• Endymion_Mallorn@kbin.melroy.org ⁨2⁩ ⁨days⁩ ago

  Honestly, it's just become my default for anything that hits the surface level "huh, that's cool...but how or why?" check nowadays. LLMs generate material that matches surface level interest without the 'why', because there's no actual thoughts or emotions to give a reason. So whenever I see a "cool" remake of a game in a different way, I start out assuming it's an LLM project driven by some slop-happy inhuman fleshbeast that doesn't understand what human elements made the old thing cool in the first place.

  So basically it's just a gut feeling that something is 'off'.

  source